Merge linux-drm-v5-6-rc3

merge linux-drm-v4-4-143

branches: 1.1.1;
Initial revision

Merge linux-drm-v5-6-rc3

merge linux-drm-v4-4-143

branches: 1.1.1;
Initial revision

drm/ttm: Avoid uninitialized mem in error branch.

Not sure why this error branch is getting hit, but let's not make the
problem worse by choking on stack garbage.

Candidate fix for symptom of PR kern/56565, PR kern/56711.
Underlying problem -- that ttm_bo_mem_space fails with ENOMEM --
remains.

Cast addr to uint64_t before printing.
Ok: riastradh@

drm: Switch sense of #if __linux__ conditionals to use __NetBSD__.

Keep it consistent with the rest of drm.

drm/ttm: Conditionalize destruction of vma node.

Needs to match conditional initialization.

Fix merge botch: do uvm_obj_init unconditionally.

drm/ttm: Omit spurious kfree in global failure branches.

drm/ttm: Fix up ttm_bo_global_release a bit.

Don't call TTM functions that were removed, try replacements


Author: Maya Rashish <maya@NetBSD.org>

do a struct mutex we can digest, use more specific header than drmP


Author: Maya Rashish <maya@NetBSD.org>

More obvious part of TTM API changes:

- ref/unref to get/put
- some global TTM things are now global
- More indirection to get vma_node


Author: Maya Rashish <maya@NetBSD.org>

Merge linux-drm-v5-6-rc3

UVM locking changes, proposed on tech-kern:

- Change the lock on uvm_object, vm_amap and vm_anon to be a RW lock.
- Break v_interlock and vmobjlock apart. v_interlock remains a mutex.
- Do partial PV list locking in the x86 pmap. Others to follow later.

Reduce diffs by side-loading some header files like Linux.

From riastradh

Revert cosmetic changes.

Use a header file hack to cope with Linux/NetBSD namespace clashes.

Currently serves for:

ALIGN
mutex_init
mutex_destroy

mark case fallthrough to avoid warning

branches: 1.15.6;
combine a pair of #ifdef's and make upstream diff slightly less ugly. NFCI.

Check for DMA-safe addresses before allowing DRM prime import.

Need <linux/fence.h> for fence stuff.

Go back from DRM_HZ to HZ, seems fine.

merge linux-drm-v4-4-143

branches: 1.10.16; 1.10.18;
Make sure mem.bus.is_iomem is initialized. PR 49833

Use KASSERTMSG to show bad base/offset.

Need the page's uvm object lock to do pmap_page_protect.

Use pmap_pv(9) to remove mappings of device pages in TTM.

Adapt nouveau and radeon to do pmap_pv_track for their device pages.

Proposed on tech-kern with no objections:

https://mail-index.netbsd.org/tech-kern/2015/03/26/msg018561.html

Further background at:

https://mail-index.netbsd.org/tech-kern/2014/07/23/msg017392.html

branches: 1.6.2; 1.6.4;
Simplify previous.

For gem and ttm objects backed by uvm_aobjs, share the vmobjlock.

XXX pullup to 7

branches: 1.4.2;
Revert ttm_bo.c 1.3 and try to fix the double-free properly.

bo->destroy doesn't ttm_tt_destroy or ttm_tt_fini anything. Trouble
is that ttm_tt_fini expects the swap storage to still be there, and
ttm_tt_destroy -- which is what calls ttm_tt_fini by way of
ttm->func->destroy -- has already nulled it out.

Don't ttm_tt_destroy bo->ttm if bo->destroy will destroy it too.

First whack at radeon driver.

No hardware to test yet, but it builds.

branches: 1.1.1;
Initial revision

Destroy lock before freeing


Author: Maya Rashish <maya@NetBSD.org>
Committer: Taylor R Campbell <riastradh@NetBSD.org>

Merge linux-drm-v5-6-rc3

Reduce diffs by side-loading some header files like Linux.

From riastradh

branches: 1.3.6;
merge linux-drm-v4-4-143

branches: 1.2.4; 1.2.28; 1.2.30;
First whack at radeon driver.

No hardware to test yet, but it builds.

branches: 1.1.1;
Initial revision

drm/ttm: Thread the needle of buffer transfer object initialization.

Kinda finicky here!

Apparently vunmap needs an argument which is num pages... one?


Author: Maya Rashish <maya@NetBSD.org>
Committer: Taylor R Campbell <riastradh@NetBSD.org>

drm/ttm: Conditionalize initialization of vma node.

Needs to match conditional destruction.


Author: Maya Rashish <maya@NetBSD.org>
Committer: Taylor R Campbell <riastradh@NetBSD.org>

drm_os_netbsd.h is no more

Follow up with TTM API changes


Author: Maya Rashish <maya@NetBSD.org>

drm_os_netbsd.h for CONFIG_THINGS. Might want a better solution...


Author: Maya Rashish <maya@NetBSD.org>

Fix up ttm.

Merge linux-drm-v5-6-rc3

UVM locking changes, proposed on tech-kern:

- Change the lock on uvm_object, vm_amap and vm_anon to be a RW lock.
- Break v_interlock and vmobjlock apart. v_interlock remains a mutex.
- Do partial PV list locking in the x86 pmap. Others to follow later.

Reduce diffs by side-loading some header files like Linux.

From riastradh

Revert cosmetic changes.

Reduce diffs by nixing DRM_SUSER macro, deleted upstream long ago.

Instead stub Linux capable(CAP_SYS_ADMIN).

Use a header file hack to cope with Linux/NetBSD namespace clashes.

Currently serves for:

ALIGN
mutex_init
mutex_destroy

ttm_io_prot: follow linux semantics and set either PMAP_WRITE_COMBINE or
PMAP_NOCACHE when existing cache flags are not set

branches: 1.14.6;
This is not merely `not for NetBSD'; upstream it's #if 0 altogether.

Always has been.

restore #ifndef __NetBSD__ code so drm on nouveau has a chance.
from @riastradh.

display is still black, but tools some times try to work now.

Be consistent about I/O byte ordering.

Leave XXX comments where it's not clear what we should do.

Reduce some logic duplication and just use vmap/kmap here.

Handle removal of local variable in merge.

Actually do DRM_SUSER check for mapping >1 page in kernel.

Need <asm/barrier.h> for mb.

merge linux-drm-v4-4-143

branches: 1.6.16; 1.6.18;
vunmap takes 2 parameters on netbsd

First attempt to make ttm_buffer_object_transfer less bogus.

branches: 1.4.2; 1.4.4; 1.4.6;
Don't forget to pmap_update after pmap_kremove!

We need just a VA, not wired physical pages too.

First whack at radeon driver.

No hardware to test yet, but it builds.

branches: 1.1.1;
Initial revision

Merge linux-drm-v5-6-rc3

merge linux-drm-v4-4-143

branches: 1.1.1;
Initial revision

Merge linux-drm-v5-6-rc3

Reduce diffs by side-loading some header files like Linux.

From riastradh

branches: 1.4.6;
Mark unused variable.

merge linux-drm-v4-4-143

branches: 1.2.4; 1.2.28; 1.2.30;
First whack at radeon driver.

No hardware to test yet, but it builds.

branches: 1.1.1;
Initial revision

Merge linux-drm-v5-6-rc3

merge linux-drm-v4-4-143

branches: 1.1.1;
Initial revision

drm/ttm: Omit spurious kfree in global failure branches.

Revert local change; drmP.h is about to be no more.

Merge linux-drm-v5-6-rc3

Reduce diffs by side-loading some header files like Linux.

From riastradh

Reduce diffs by nixing DRM_SUSER macro, deleted upstream long ago.

Instead stub Linux capable(CAP_SYS_ADMIN).

branches: 1.4.6;
Destroy glob->lock if we init it.

merge linux-drm-v4-4-143

branches: 1.2.4; 1.2.28; 1.2.30;
First whack at radeon driver.

No hardware to test yet, but it builds.

branches: 1.1.1;
Initial revision

Merge linux-drm-v5-6-rc3

merge linux-drm-v4-4-143

branches: 1.1.1;
Initial revision

Merge linux-drm-v5-6-rc3

merge linux-drm-v4-4-143

branches: 1.1.1;
Initial revision

Merge linux-drm-v5-6-rc3

merge linux-drm-v4-4-143

branches: 1.1.1;
Initial revision

Merge linux-drm-v5-6-rc3

merge linux-drm-v4-4-143

branches: 1.1.1;
Initial revision

drm/ttm: Fix potential leak in error branch.

Not likely but let's not leave a rake to trip on.

drm/ttm: Fix ttm_tt pages and DMA map allocation.

drm: Nix use of uvm pglist. Just use arrays of page pointers.

drm/ttm: uao_create takes pages, not bytes.

drm/drm_mem_util.h for drm_free_large


Author: Maya Rashish <maya@NetBSD.org>

Fix up ttm.

Merge linux-drm-v5-6-rc3

Reduce diffs by side-loading some header files like Linux.

From riastradh

branches: 1.11.6;
merge linux-drm-v4-4-143

branches: 1.10.2; 1.10.4;
Fix error branch in ttm_dma_tt_init to avoid double-free.

Should fix symptom of PR kern/52438, but who knows what underlying
problem causes us to reach the error branch in the first place.

branches: 1.9.6;
Clarify ttm state transitions tt_unpopulated<->tt_unbound<->tt_bound.
Assert it, too, and don't handle other cases.

We can add the assertion to ttm_agp_tt_unpopulate because it is called by
{nouveau,radeon}_ttm_tt_unpopulate, which is generically called
ttm_tt_unpopulate.

And the sole caller to ttm_tt_unpopulate (ttm_tt_destroy) only does so if the
state is unbound. the other caller is in a !NetBSD block.

We can add the assertion to ttm_agp_tt_populate and avoid handling the
!unpopulated case because the sole callers are {nouveau,radeon}_ttm_tt_populate
both of which return early in the !unpopulated case.

We can change the assertion on ttm_tt_wire because it is solely called by
ttm_bus_dma_populate, which already asserts that it is the unpopulated case.

from riastradh

branches: 1.8.2;
Guarantee no zero-size uao/kmem allocations via ttm.

It may be that all callers guarantee no zero-size ttm objects, but I
can't prove that in five minutes of browsing callers. Rather than
add a KASSERT, lacking proof, we'll add a warning message so that if
it does happen then it happens noisily, but we'll also prevent the
bad consequences of passing zero into uao_create by rounding up to a
harmless nonzero allocation.

XXX pullup-7

branches: 1.7.2;
Rework ttm tt swapin/swapout logic.

Rather than handling `swapping in/out' here, per se, we let uvm do
that, we interpret `swap out' as `deactivate pages', and we add
generic ttm operations to wire and unwire pages, for the ttm_tt
driver to use.

This fixes certain graphics buffer eviction logic, which enables
nouveau to suspend/resume on one of my machines. (The machine
doesn't resume overall for other reasons, but the nouveau device
suspends and resumes in isolation.)

XXX pullup to netbsd-7 after a couple weeks

branches: 1.6.2; 1.6.4; 1.6.6; 1.6.8; 1.6.10;
Revert ttm_bo.c 1.3 and try to fix the double-free properly.

bo->destroy doesn't ttm_tt_destroy or ttm_tt_fini anything. Trouble
is that ttm_tt_fini expects the swap storage to still be there, and
ttm_tt_destroy -- which is what calls ttm_tt_fini by way of
ttm->func->destroy -- has already nulled it out.

Fix non debug build.

Fix ttm swap storage destruction.

Don't forget to advance the array index as we walk the queue...

First whack at radeon driver.

No hardware to test yet, but it builds.

branches: 1.1.1;
Initial revision